Fetal Health and the Environment

Abstract

Whether fetuses are vulnerable to the effects of environmental exposures is a difficult question to answer due to the internal, and therefore, hidden nature of conceptions and early embryonic development. Studies of prenatal environmental exposures often assess fetal health using newborn anthropometry, such as birth weight, as these data are easily accessible from medical records and birth certificates. Although newborn size is a key development indicator for perinatal morbidity and mortality under the Barker hypothesis (also known as the Developmental Origins of Health and Disease hypothesis), it is measured at the end of pregnancy, and thus provides limited insights on the timing of when the growth-restricting effects manifest. That is, birth weight is insensitive to early to mid-pregnancy effects, as a fetus that experiences early fetal growth restriction can still catch up to achieve population growth standards by birth. Furthermore, birth weight is mainly driven by fat accumulation throughout pregnancy, and so may be a poor proxy for the development of other organs relevant for health (e.g., it may not reflect a reduction in head size, which has implications for later brain development). Finally, assessing newborn health necessitates the production of live births. Most epidemiologic analyses have focused on birth outcomes, and so ignore the effects on pregnancy loss, which not only is a relevant health outcome, but also a potential source of bias, as it is a competing event that prevents the birth outcome of interest from occurring. Thus, this dissertation comprises of three studies that aim to examine how environmental exposures affect in utero fetal developmental processes during pregnancy, rather than using proxies at birth. In Chapter 1, we assessed through simulations the potential bias induced by restricting epidemiologic analyses to live births when pregnancy loss is influenced by the environmental exposure of interest A and any unmeasured factors U that also affect the child outcome. Few simulation studies have explored this topic, but they presuppose that there is no interaction between the exposure A and unmeasured factor U, which may not be a realistic assumption. In this chapter, we relax these assumptions and consider three fetal survival (or selection) mechanisms: 1) collider-stratification without interaction, where A and U independently affect selection, 2) depletion of susceptibles, where selection is dependent on the presence of both A and U (i.e., the interaction-only effect), and 3) when both mechanisms operate simultaneously. We show that ignoring pregnancy loss when estimating the effects of prenatal exposures on outcomes in live-born children lead to associations that are biased downwards, where the magnitude of the bias is determined by the selection mechanism, strength of selection, and prevalence of U. In Chapters 2 and 3, we examine the association between gestational exposure to PM2.5 and ambient temperature, respectively, and fetal growth outcomes in a pregnancy cohort using spatiotemporally resolved data on exposures in Massachusetts, USA. Unlike prior studies that have mostly examined fetal growth using newborn size, we leveraged data from routine ultrasound measurements which allowed us to observe the developmental processes of distinct organs during pregnancy. These last two chapters demonstrate that gestational exposure to PM2.5 and ambient temperature were associated with impaired fetal growth, where early to mid-pregnancy appears to be a critical window of exposure. Overall, the findings from this collection show the importance of studying in utero fetal health during pregnancy, rather than at its conclusion.